Richard,

The article does propose a practical tolerance. I quote the relevant
paragraph below.

"Of course, nothing is perfect so an acceptable tolerance for depending on
the assumed Zs=50+j0? for Mismatch Loss calculations might be that ‘forward
power’ doesn’t vary by more than 10% of the ‘reflected power’ at any point.
The worst case ‘reflected power’ for the experiment as described is 4%, so
a variation of ‘forward power’ by more than 0.4% (ie 0.4W in 100W) at any
point would indicate significant error in any Mismatch calculations based
on an assumed Zs=50+j0?."

I explained in the articles how I am able to reliably detect such small
relative power variation. In one of the tests, the measured variation was
21%.

My intention in publishing the two articles wasn't to prove the propostion
one way or the other, but to show practical amateurs a test that they could
make with relatively simple equipment, one that is relevant to the
application of Zs if it was known, and an explanation of why it works so
that they can understand the test and make their own interpretations about
significance.

Owen